Last modified: 2014-06-30 18:08:56 UTC
A gadget thats crucial for the community and enabled by default was broken in Firefox and got fixed yesterday: https://www.wikidata.org/wiki/Special:Diff/129233461 It works in debug mode: https://www.wikidata.org/wiki/Q1523127?debug=1 (Note the web links in the values, e.g. the VIAF identifier is linked to the viaf.org website.) It constantly refuses to work in non-debug mode: https://www.wikidata.org/wiki/Q1523127 I tried everything I can think of: Dump the Firefox cache, dump localStore explicitly, disable the Firefox cache completely, purge both the gadget and the example page. It's reproducible on other machines. Shift+reloading the page produces random-like results: Sometimes it works a single time but if you reload the page again it's broken again. As far as I can tell either the RessourceLoader itself or some configuration is to blame. I don't know who to ask or what to do next. Please help.
Lydia pointed me to bug #65375. The two may be connected somehow but the effect is very different. The issue here is pretty much the same I reported and discussed many, many times since two or three years (#41130, #49362, #56874): Some cache (Squid, Varnish, no idea) or load balancer continues to deliver outdated garbage for days, sometimes for months. I can actually see this in the JavaScript console: It still shows an error message that was fixed in Gadget-Search.js days ago (and to repeat, yes, I'm 100% sure it's not the client). Even after years reporting and discussing this issue I still don't know a way to make the caching layers stop delivering broken content. Apparently there is none. Except for telling all users that they should switch to ?debug=1 mode? Or renaming all gadgets every time they are changed?
If I do localStore.clear() and Shift+reload the page, it works a single time on that single page. If I go to an other page or simply reload the current one it stops working (which means I get outdated garbage from some caching layer, again).
Is this still a problem? (Plus wondering if this problem is a "blocker" for something.)
(In reply to Andre Klapper from comment #3) > Is this still a problem? I don't think so... if this reoccurs, we might reopen (also there's no way to debug this right now).
(In reply to Andre Klapper from comment #3) > Is this still a problem? As always it magically faded away and nobody knows what happened. So we have the guarantee it will happen again. Which leaves me unhappy. :-( I later noticed that parts of the "random like" behavior I described above was causes by the fucked up caching in Firefox and/or the ResourceLoader dumping stuff in an unusual place that fucks up the browsers caching somehow. Still this doesn't explain everything, like being able to reproduce it on other machines and in other browsers.
Still sounds like nobody knows if it still happens. Setting to WORKSFORME, please reopen if there is any indication that we can actually investigate then when we have a testcase.
